Understanding credit risk in Norwegian real estate crowdlending : Analysis of credit quality among Norwegian real estate crowdlending borrowers across FundingPartner, Kameo and Monio
MetadataShow full item record
- Master Thesis 
The Norwegian crowdlending industry has grown rapidly in the last decade, resulting in the emergence of several platforms of notable sizes. Regulations are lagging, and government instances are discussing incorporating EU directives. This thesis aims to investigate risk differences in credit classifications across Norwegian crowdlending platforms. We identify risk factors and analyze potential differences in risk related to loans issued by FundingPartner, Kameo and Monio. We analyzed differences both for the platforms overall and within the credit classifications. The results provide an overview of differences in credit assessment that may benefit the decisions of both lenders and policymakers. The analysis is based on a manually assembled data set containing loan data, financial statements and policy rates. Our empirical analysis uses three bankruptcy models to evaluate borrowers' credit risk based on financial statements. The results from the bankruptcy models are tested to ensure significance. Moreover, we integrate project-specific risk elements such as collateral, loan size, loan term and interest rates to explain the differences we discovered. We also consider actual default rates and check if they are consistent with our empirical results. Despite having equal credit classification, we discovered significant differences between borrowers of such loans. FundingPartner issued A-classified loans with significantly riskier borrowers than Monio, despite Monio rewarding their lenders with higher interest rates. Borrowers of Monio are overall the least risky, yet the platform hosts the riskiest borrowers in our sample. Kameo borrowers with D-classified loans are significantly less risky than Monio's. Furthermore, we observe considerable differences in the use of collateral to secure lenders in the event of default. Lastly, we compare our empirical findings against confirmed defaults.